OBJECTIVE: We conducted a study to identify gender differences in factors associated with the first episode of non-adherence in the 12 months following the first antiretroviral prescription. METHODS: A concurrent prospective study of patients initiating antiretroviral therapy in Brazil was conducted from 2001-2002. The self-reported measurement of adherence was defined as an intake of less than 95% of the prescribed number of doses. Only the first occurrence of non-adherence was considered in this analysis. All analyses were stratified by gender. A Cox proportional hazard model was used to estimate the risk of non-adherence, and the time to non-adherence was estimated using the Kaplan-Meier method. RESULTS: The cumulative incidence of non-adherence was 34.6% (29.7% and 43.9% among men and women, respectively; p=0.010). Marital status (being married or in stable union; p=0.022), alcohol use in the month prior to the baseline interview (p=0.046), and current tobacco use (p=0.005) increased the risk of non-adherence among female participants only, whereas a self-reported difficulty with the antiretroviral treatment was associated with non-adherence in men only. For both men and women, we found that a longer time between the HIV test and first antiretroviral therapy prescription (p=0.028) also presented an increased risk of non-adherence. CONCLUSIONS: In this cohort study, the incidence of non-adherence was 1.5 times greater among women compared to men. Our results reinforce the need to develop interventions that account for gender differences in public referral centers. Additionally, we emphasize that, to achieve and maintain appropriate adherence levels, it is important to understand the barriers to seeking and utilizing health care services.